Prospective cohort study of prone sleeping position and sudden infant death syndrome

Studies of the link between prone sleeping position and sudden infant death syndrome have been criticised on grounds of recall bias and for not taking into account possible confounding effects. To avoid recall bias and to allow measurement of important biological factors a prospective cohort study of the cause of sudden infant death syndrome (SIDS) is being conducted. The infants included are those at high risk of the syndrome as assessed by a perinatal score. Of the 3110 members of the cohort born between January, 1988, and end of March, 1990, 23 infants later died of SI DS. Sleep position information was available for 15 of these. Matched analysis to control for the confounding effects of infant birthweight and maternal age indicated that prone sleeping position was associated with an increased risk of SIDS (OR 4·47 95% Cl [1·30-15·43]). The findings are strengthened by the results of a concurrent retrospective case-control study of 42 SIDS cases in which the prone position was also associated with an increased risk of SIDS (unadjusted OR 3·45 [1·59-7·49]).

Here we present data from a prospective study on the relation between SIDS and prone sleeping position. Several likely confounders are controlled for. In addition, the strength of association between SIDS and prone or non-prone position on the basis of prospectively collected data was compared with that found by retrospective data collected on the same infants.

Methods Tasmania, the island state of Australia, has approximately 7000 livebirths per annum, and its rate of SIDS-3.5 per 1000 livebirths-is considerably higher than that in other Australian states.19 A prospective cohort study was started in January, 1988, to investigate the cause of SIDS in Tasmania. The six obstetric hospitals taking part cover approximately 93% of livebirths in the state. Infants born within these hospitals are assessed according to a local scoring system20 to predict infants at high risk of SIDS. The composite score is based on maternal age, birthweight, season of birth, sex, duration of the second stage of labour, and infant feeding. Infants with a score over a cut-off point2O are eligible for the study. The cut-off point identifies a group which represents approximately one fifth of livebirths in the state. Multiple births are also included in the study. Infants with severe neonatal disease or a major congenital anomaly and infants for adoption are excluded from the study.

Prospective cohort study Introduction

Despite extensive research, sudden infant death syndrome (SIDS) remains a major determinant of infant mortality in developed countries. The apparent lack of progress in this disorder is due to the failure to find, in human studies, consistent strong associations that might be putative risk factors and that might readily be altered by public health measures. However, there is increasing evidence that the prone sleeping position is related to the risk of sudden infant death syndrome. The prone sleeping position has been found to be commoner among cases than controls in retrospective work, although this relation has not always been statistically significant, with odds ratios for prone sleeping position and SIDS varying from 17 to 12-5.1-l In communities where infants rarely sleep prone, sudden infant death syndrome is uncommon.2,12 Clearly such findings have important implications for prevention of SIDS. Already, in some locations these findings have stimulated a change in the way babies are positioned, with-in South Australia and the Netherlands, for example-a concomitant decline in SIDS incidence.13,14 Nevertheless, these observations and the inference that prone position might be an important causal factor in SIDS have not been received uncritically. Two important issues have been raised. The first issue is that of recall bias in the interview of bereaved parents;"’ prospectively collected information is required to establish the validity of the data on risk factors such as prone position in sudden infant death.1s The second concern is that potentially important confounders, such as low birthweight and maternal education,18 have not been taken into account in much of the previous work.

From January, 1988, until March, 1990,3110 infants (20% of all livebirths) were eligible for the survey. 2977 (96%) of these eligible infants participated in the hospital interview and 2607 (84%) participated in both the hospital and home interviews. Data are obtained by research assistants on three occasions. At the hospital interview conducted on day 4 of life, sociodemographic, obstetric, and perinatal data (including information on maternal nutrition and alcohol and smoking practice during pregnancy), birthweight, height, head circumference, and triceps and subscapular skinfold thicknesses are recorded. The second occasion is a home visit conducted during the fifth postnatal week, but for premature infants ( < 36 weeks’ gestation) this visit is done at 40 weeks post conceptional age. For babies kept in hospital because of complications the home visit is delayed until the infant has been at home for at least 2 weeks. At the home visit a comprehensive questionnaire is used to elicit information on usual sleep position, usual sleep pattern, body movement during sleep, infant illness and health service attendance, pattern and type of infant feeding, parental smoking practice, child’s overnight clothing and bedding, home heating and housing, and other characteristics of the infant and of parental care. Anthropometric and temperature measurements are made on the infant and a developmental assessment is conducted. The physical features of the house are noted and humidity and 24’ hour max-min infant bedroom temperature on day of home visit is recorded. The third occasion is a phone interview conducted when the infant is 10 weeks old to review infant progress, illness history, feeding, and immunisation. To collect information on sleeping position mothers are asked verbally "What position does your baby usually sleep in?". The prone position refers to the infant positions "on stomach, face to ADDRESSES: Menzies Centre for Population Health Research, University of Tasmania, 43 Collins Street, Hobart, Tasmania, Australia 7000 (Prof T Dwyer MD, A-L B. Ponsonby, MB, N M Newman, FRCP, L. E Gibbons, MS). Correspondence to ProfT Dwyer



and controls. The same odds ratios were also calculated for the entire retrospective case-control study.


Age at death distribution of the cohort members who later died of SIDS.

side" and "on stomach, face down", whereas "on side", "on back, face to the side", "on back, face up", and "other" are classified as non-prone.

Retrospective case-control study A retrospective case-control study has been conducted since October, 1988, with cases being all infants dying suddenly and unexpectedly in Tasmania. 2 matched controls are selected for each case, control A being matched for hospital of birth and postnatal age, and control B for hospital of birth, postnatal age, and birthweight (< 1500 g, 1500-2500 g, > 2500 g). Case information on usual sleeping position was collected by an interviewer, who administered a questionnaire to parents 5 weeks after the death of an

infant. Control infants are interviewed within a month of the index case’s death, and they are chosen to be the same postnatal age at interview as the SIDS infant was at death. Information on usual sleeping position is included in the data recorded. From October, 1988, until the end of June, 1990, 42 of the 43 SIDS cases occurring during this interval were entered into the study. 83% of the controls initially chosen were interviewed. If a control infant was unable to be interviewed another appropriate infant was chosen. All infants dying suddenly and unexpectedly in Tasmania during both study periods underwent a postmortem examination by a hospital pathologist. Toxicological and bacteriological studies were done routinely.

Statistical methods

Prospective data-The relative risk for prone sleeping position and SIDS was estimated for the entire birth cohort. To identify potential confounders data were subjected to univariate relative risk analysis.21 The individual effect of the identified confounders was assessed by traditional stratified analysis.22 In the third stage of the analysis a "nested" case-control analysis was done, with controls selected from the non-SIDS population of the cohort for each case. The controls were matched on birthweight, maternal age, and availability for interviews, and were born within 1 month of the case. 4-10 controls were selected for each case, depending on available matches. The matching of the cases and controls was preserved in the analysis. Multiple conditional logistic regression22 was used on this prospectively collected case-control data to obtain an odds ratio for prone sleeping position and SIDS and to assess additional variables as potential confounders. The odds ratio and 95% confidence interval from the conditional logistic regression are logit-based.22 All other odds ratios and relative risks use MantelHaenszel estimates with test-based 95% confidence intervals.23 Retrospective data.-Parents of 13 of the 15 SIDS cases on whom information on sleeping position was collected at 1 month of age in the cohort study answered the retrospective questionnaire section of the case-control study. Unadjusted odds ratios for usual sleeping position and SIDS outcome were calculated for these cases

Prospective cohort study Among the 3110 infants born between January, 1988, and March, 1990, and who met the inclusion criteria, 23 infants had later died suddenly and unexpectedly, with no cause being identified at necropsy, as of June, 1990. The rate of SIDS for infants in the cohort study was 7-4 per 1000 livebirths. The age-at-death distribution for SIDS cases is outlined in the figure. 3 SIDS infants died before home visit age. Home information was obtained prospectively for 15 of the 20 eligible SIDS infants, giving a response rate of 75%. The overall response rate for the entire cohort was 84%. The 23 SIDS deaths represent 46% of all infants dying from SIDS in Tasmania who were born between January, 1988, and March, 1990, and who died before the end of June, 1990. The distribution of infant sleeping position is given in table I. The relative risk for usual prone sleeping position at 1 month of age and SIDS is 3-12 (95% CL 1-18, 8-29). Of the variables examined, infant birthweight, maternal age, and maternal smoking habit were found to be likely confounders of the relation between prone sleeping position and SIDS (table 11). The individual effect on each of these factors was then assessed by stratified analysis. In the study sample 45 % of low birthweight (< 2500 g) infants and 29 % of normal birthweight ( 2500 g) infants were sleeping in the prone sleeping position at 1 month of age. The adjusted relative risk for the prone sleeping position and SIDS was 2-69 (1-05, 6-94) after stratification by birthweight; 3-43 (1-29, 9-15) after stratification by maternal age ( < 20 yr, 20 yr); and 3-58 (1-21, 10-64) after stratification by maternal postnatal smoking habit (non-smoker, smoker). Maternal age and smoking habit were found to be closely related and, unlike birthweight, maternal smoking habit became non significant after controlling for maternal age. For the multiple regression procedure, controls were matched on maternal age and birthweight, as well as on TABLE II-RELATIVE RISKS FOR SIDS AND PRONE SLEEPING POSITION FOR SELECTED STUDY VARIABLES AMONG ALL





*Numbers m parentheses refer to number of cases tControls from the prospective study and retrospective study as described


the text




*Control infants matched on birthweight, maternal age, availability for interviews and born within one month of case. tAs recorded at the 1 -month interview

availability for interviews and date of birth. Three of the variables used in selecting the cohort-infant sex, infant feeding practice, and duration of second stage labour-were not used as matching variables in selecting cases for this analysis. However, these three variables were found not to affect the relation between prone sleeping position and SIDS. A total of 116 controls were chosen for the 15 cases (table ill). A comparison of the distribution of selected variables between cases and controls chosen from the cohort study is shown in table ill. Multivariate analysis was used to test the effects of the

following possible confounders: infant post-conceptional or postnatal age at interview, maternal age, maternal cigarette smoking, infant sex, length of gestation, birthweight, cot type, mattress lining and type, pillow type, heater use in infant’s room, breast feeding, breathing and feeding problems, history of an upper respiratory tract infection, perspiring when not febrile, visiting a doctor, developmental prone posture, sleep movement, birth as a public hospital patient, region of Tasmania, family income, and parents’ education. None of these factors was a confounder of the relation between prone position and SIDS. In addition, the association between prone sleeping position and SIDS did not vary by season. Therefore, in the matched analysis on the "nested" case-control data only birthweight, maternal age, availability for interview, and month of birth were held constant. In this multivariate analysis the odds ratio for prone sleeping position at 1 month of age and SIDS is 4 47

(130, 1543).

Comparison of prospective and retrospective data on sleeping position For infants for whom both prospectively and retrospectively collected data on usual sleeping position were available, there was close agreement between both sets of data (table iv). Prone position was significantly associated with an increased risk of SIDS in both the cohort and case-control study (p

Prospective cohort study of prone sleeping position and sudden infant death syndrome.

Studies of the link between prone sleeping position and sudden infant death syndrome have been criticised on grounds of recall bias and for not taking...
602KB Sizes 0 Downloads 0 Views